Method and apparatus for measuring paper fibers

Abstract

A paper fiber measurement method and a paper fiber measurement device, the paper fiber measurement method comprising the following sequentially executed steps: a defect area removal step, removing a defect area in a to-be-detected image and saving a defect-free effective image; an image partition cutting step, dividing the defect-free effective image into a plurality of non-overlapping sub-areas and recording effective pixel areas of the sub-areas; a partition weighting analysis step, calculating weights according to the effective pixel areas of the sub-areas, and performing multi-dimensional uniformity index analysis on each sub-area; and a comprehensive weighting result step, performing weighted calculation on multi-dimensional uniformity indexes of the sub-areas based on the weights, and outputting comprehensive detection data of fiber distribution uniformity. The paper fiber measurement method and the paper fiber measurement device can remove the defect area in the paper image, accurately output the fiber distribution uniformity index through the partition area weighting and multi-dimensional quantitative analysis, and improve the detection precision and reliability.

Description

Technical Field

[0001] This invention relates to a paper fiber measurement method and a paper fiber measurement device, and more specifically, to a paper fiber measurement method and a paper fiber measurement device capable of quantitatively detecting the uniformity of paper fiber distribution based on computer vision algorithms. Background Technology

[0002] In the papermaking industry, the uniformity of paper fiber distribution is a key physical indicator that determines product quality. It directly affects the paper's mechanical strength, printability, light transmittance consistency, and other key properties, and plays a decisive role in the performance of downstream applications such as printing paper, packaging paper, and specialty industrial paper.

[0003] As papermaking processes become faster and wider, and as the market demands for paper quality stability, traditional uniformity testing methods are no longer sufficient to meet the quality control needs of modern production. Computer vision inspection technology, with its advantages of being non-contact, highly efficient, and quantifiable, is gradually becoming the mainstream technology in this field.

[0004] Furthermore, with the development of computer vision and image processing technologies, machine vision-based detection systems have been widely used in the paper industry, enabling efficient identification and location of common paper defects such as holes, dust, scratches, and pulp stains on the paper surface.

[0005] Currently, mainstream paper defect detection models use deep learning algorithms to collect images and identify paper defects in real time during the paper production process. The detection speed can reach several meters per second, significantly improving the defect detection rate and effectively reducing labor costs and scrap rates.

[0006] However, existing computer vision-based detection technologies still have shortcomings in the quantitative analysis of fiber distribution uniformity, with the main problem being the processing of images containing defects.

[0007] In actual production, the images collected by the paper defect detection model inevitably contain various obvious defects, such as holes, stains, wrinkles, and fiber agglomerates. The gray values ​​and texture features of these defective areas are fundamentally different from those of normal fiber distribution areas. If images containing defects are directly used for uniformity analysis, it will lead to serious distortion of key quantitative indicators such as gray mean and standard deviation.

[0008] For example, the extremely low grayscale value in the pore area due to the lack of fiber obstruction will lower the overall grayscale mean and increase the standard deviation, causing the detection system to misjudge it as uneven fiber distribution; while defects such as pulp spots and stains will cause abnormally high local grayscale values, which will also interfere with the accuracy of the uniformity index, ultimately leading to quality judgment deviation and thus affecting the effectiveness of production process adjustments.

[0009] In view of the shortcomings of the existing technology, how to provide a paper fiber measurement method and paper fiber measurement device that can improve the detection accuracy of paper fiber distribution uniformity has become an urgent technical problem to be solved. Summary of the Invention

[0010] This disclosure is made to solve the above-mentioned technical problems, and its purpose is to provide a paper fiber measurement method and a paper fiber measurement device. The paper fiber measurement method and the paper fiber measurement device can remove defective areas in paper images, and accurately output fiber distribution uniformity indicators through partitioned area weighting and multi-dimensional quantitative analysis, thereby improving detection accuracy and reliability.

[0011] To achieve the objectives of this disclosure, a paper fiber measurement method is provided for quantitatively detecting the uniformity of paper fiber distribution, comprising the following steps performed sequentially: Step 1: Defect region removal, acquiring an image of the paper to be detected, removing defective regions from the paper image and saving a defect-free valid image; Step 2: Image partitioning, dividing the aforementioned defect-free valid image into multiple non-overlapping sub-regions and recording the effective pixel area of ​​each sub-region; Step 3: Partition weighted analysis, calculating weights based on the effective pixel area of ​​each sub-region, and performing multi-dimensional uniformity index analysis on each sub-region; Step 4: Comprehensive weighted result, performing weighted calculation on the multi-dimensional uniformity index of each sub-region based on the aforementioned weights, and outputting comprehensive fiber distribution uniformity detection data.

[0012] Based on the above-described structure, this method effectively solves the technical problems of defect interference and one-sided analysis in traditional detection through "defect removal, partitioning, weighted analysis, and comprehensive calculation". It realizes standardized quantitative analysis of the entire process from image preprocessing to index output, and significantly improves the accuracy and reliability of fiber distribution uniformity detection.

[0013] Preferably, in step 1 above, the user's instruction to select the defective area in the original paper image is received through an interactive operation interface, the defective area is separated from the original image and marked with a box, and the effective image after removing the defect is saved simultaneously.

[0014] Based on the above-mentioned structure, the defect removal method, which combines interactive selection with visual labels, not only ensures the accuracy of defect area location, but also clearly excludes ineffective analysis areas through intuitive coverage labels, effectively avoiding distortion of indicators such as grayscale mean and standard deviation caused by defects.

[0015] Preferably, in step 2 above, based on the pixel coordinate system, the effective area is divided into four sub-regions—upper, lower, left, and right—along the horizontal and vertical central axes, using the smallest bounding rectangle as the reference.

[0016] Based on the above-described configuration, the central axis partitioning method based on pixel coordinates ensures the regularity and non-overlapping nature of the sub-region division, achieving comprehensive coverage analysis of the effective area of ​​the paper and effectively avoiding misjudgment of uniformity caused by omission of local areas.

[0017] Preferably, in step 3 above, the weights are calculated using the following mathematical formula (1), where Weight i Area represents the weight of the i-th partition. i Let ∑Area represent the effective pixel area of ​​the i-th partition. all parts This represents the total effective pixel area of ​​all sub-regions.

[0018] Based on the above-mentioned structure, the effective pixel area ratio of the sub-region is used as the weight, so that the sub-region with a larger area occupies a more reasonable weight ratio in the comprehensive analysis, avoiding the analysis bias caused by the difference in the area of ​​each sub-region, and ensuring that the weighted result can truly reflect the fiber distribution characteristics of the whole paper.

[0019] Preferably, in step 3 above, the multi-dimensional uniformity index includes the average grayscale value, which is calculated using the following mathematical formula (2), wherein, This represents the average grayscale value of the i-th partition. Let be the pixel height (row number) of the i-th partition. Let be the pixel width (number of columns) of the i-th partition. This represents the total pixel area of ​​the partition. This represents the grayscale value of the pixel at coordinates (x, y) within the i-th partition.

[0020] Based on the above-mentioned structure, by accurately summing and averaging pixel gray values, the overall brightness of each sub-region can be objectively reflected, indirectly characterizing the fiber packing density. This effectively avoids interference from single pixel anomalies on the overall features of the region and improves the reliability of density-related indicators.

[0021] Preferably, in step 3 above, the multi-dimensional uniformity index includes the gray-level standard deviation, which is calculated using the following mathematical formula (3), wherein, This represents the global uniformity (grayscale standard deviation) of the i-th partition. Let be the pixel height (row number) of the i-th partition. Let be the pixel width (number of columns) of the i-th partition. This represents the total pixel area of ​​the partition. This represents the pixel grayscale value at coordinates (x, y) within the i-th partition. Let be the average gray value of the i-th partition.

[0022] Based on the above structure, by quantifying the dispersion of gray values, the uniformity of fiber distribution within a sub-region is directly reflected. The smaller the standard deviation, the more regular the fiber density distribution. This index can complement the gray value average.

[0023] Preferably, in step 3 above, the multi-dimensional uniformity index includes the mean gray-level gradient, which is calculated using the following mathematical formula (4): where, This represents the mean gradient of the i-th partition. Let be the pixel height (row number) of the i-th partition. Let be the pixel width (number of columns) of the i-th partition. G(x,y) represents the total pixel area of ​​the partition, and G(x,y) represents the gradient value of each pixel.

[0024] Based on the above structure, by calculating the average value of pixel grayscale change intensity, it is possible to accurately capture detailed features such as fiber edges and texture transitions. The smaller the gradient mean, the more regular the fiber arrangement and the smoother the texture change. This index effectively makes up for the deficiencies of grayscale mean and standard deviation in texture feature characterization.

[0025] Preferably, in step 3 above, the multi-dimensional uniformity index further includes a grayscale standard deviation index calculated based on the grayscale standard deviation and the grayscale average value, and a grayscale gradient index calculated based on the grayscale gradient mean and the grayscale standard deviation.

[0026] Based on the above-mentioned structure, the grayscale standard deviation index eliminates the interference caused by the difference in fiber density in different zones, making the index more sensitive to uniformity differences; the grayscale gradient index highlights the difference in texture consistency, can effectively identify local anomalies such as fiber agglomeration, and further improves the ability to distinguish complex fiber distribution states.

[0027] Preferably, in step 4 above, the comprehensive weighted calculation includes at least one of the following: the overall global grayscale average value, grayscale standard deviation, grayscale gradient mean, grayscale standard deviation index, and grayscale gradient index.

[0028] Based on the above-mentioned structure, a comprehensive global index is output by weighted integration of multi-dimensional sub-region indicators. The overall uniformity is accurately quantified through weight balancing. Different comprehensive indicators reflect the paper fiber distribution characteristics from different dimensions, significantly improving the practicality and comprehensiveness of the test results.

[0029] Preferably, the methods for obtaining the above-mentioned paper images include: mobile phone shooting, high-speed camera shooting, and online defect monitoring camera shooting.

[0030] Based on the above-described configuration, this measurement method is compatible with various application scenarios such as offline sampling inspection, on-site rapid verification, and online detection on the production line. It can meet the lightweight requirements for convenient data collection on mobile terminals, and also adapt to the high-speed, wide-width continuous data collection requirements of paper production lines, effectively ensuring the accuracy and versatility of the entire process detection.

[0031] Furthermore, the present invention also provides a paper fiber measuring device for implementing the paper fiber measuring method described above. The paper fiber measuring device includes: a defect region removal unit, which performs defect region removal operations and saves defect-free valid images; an image partitioning and cutting unit, which performs image partitioning and cutting operations and records the effective pixel area of ​​each sub-region; a partition weighted analysis unit, which calculates the weight of each sub-region and performs multi-dimensional uniformity index analysis; and a comprehensive weighted result unit, which performs weighted calculations on the multi-dimensional uniformity index and outputs a detection report.

[0032] Preferably, the paper fiber measuring device integrates a deep learning-based paper defect detection model, which can automatically identify defective areas and mark their coordinates.

[0033] Based on the above-described structure, the deep learning model enables automated defect identification and localization without the need for manual selection, significantly improving the efficiency and accuracy of defect removal, reducing labor costs, and ensuring the consistency of defect identification while avoiding human judgment bias.

[0034] Preferably, the aforementioned paper fiber measuring device is based on a web-based front-end interface, adapted to mobile terminals, and provides a partition cutting function with selectable fast and precise modes.

[0035] Based on the above-described structure, the web-based front-end design enables multi-terminal adaptation without the need to install a dedicated client, meeting the requirements of scenarios such as offline sampling inspection and rapid on-site verification. It balances detection efficiency and accuracy, expanding the application scenarios and practicality of the device. Attached Figure Description

[0036] In view of the above objectives, the technical features of the present invention are clearly described in the following technical solutions, and its advantages are apparent from the following detailed description with reference to the accompanying drawings, which illustrate preferred embodiments of the invention by way of example, without limiting the scope of the inventive concept.

[0037] Figure 1 This is a flowchart illustrating the paper fiber measurement method of the present invention.

[0038] Figures 2A to 2G This is a schematic diagram illustrating the rejection and cutting steps of the paper fiber measurement method of the present invention.

[0039] Figures 3A to 3B This is a schematic diagram illustrating the calculation steps of the paper fiber measurement method of the present invention.

[0040] Figure 4A This is a schematic diagram illustrating samples of multiple embodiments. Figure 4B This is a comparison table of the paper fiber measurement method of the present invention and parameters obtained by laboratory measurement, for various embodiments. Figure 4C This is a schematic diagram showing the comparison curves of the two. Figure 4D This is a schematic diagram showing the fitted curves of the two. Detailed Implementation

[0041] Various embodiments of the present invention will now be described in detail, examples of which are shown in the accompanying drawings.

[0042] Although the invention has been described in conjunction with exemplary embodiments, it should be understood that this specification is not intended to limit the invention to the exemplary embodiments described below. Rather, the invention is intended to cover not only these exemplary embodiments, but also various alternatives, modifications, equivalents, and other embodiments that may be included within the spirit and scope of the invention as defined by the appended claims.

[0043] Unless otherwise defined, the technical or scientific terms used in the claims and description shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in the patent application description and claims of this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. The terms "an" or "a" and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms "comprising" or "including" and similar terms mean that the element or object preceding "comprising" or "including" encompasses the element or object listed following "comprising" or "including" and its equivalents, and do not exclude other elements or objects. The terms "connected" or "linked" and similar terms are not limited to physical or mechanical connections, nor are they limited to direct or indirect connections.

[0044] Furthermore, in the following description, in order to clearly demonstrate the structure and operation of the present invention, a number of directional terms will be used. However, terms such as "front", "back", "left", "right", "outside", "inside", "outward", "inward", "up", and "down" should be understood as convenient terms and not as limiting terms.

[0045] The paper fiber measurement method of the present invention is applicable to online inspection scenarios in paper production lines. It achieves quantitative analysis of the uniformity of paper fiber distribution through computer vision technology. Its core is to build an image processing system based on OpenCV (an open-source computer vision library), which can accurately remove defect interference and output objective quantitative indicators through weighted analysis. It can adapt to various application scenarios such as high-speed production and offline inspection.

[0046] Hereinafter, with reference to the accompanying drawings, the paper fiber measurement method of the present invention will be briefly described.

[0047] like Figure 1 As shown, the paper fiber measurement method of the present invention mainly includes four sequentially executed steps: defect area removal, image partitioning, partition weighted analysis, and comprehensive weighted result. Through the synergistic effect of each step, the accuracy and reliability of uniformity analysis can be ensured. The specific implementation process of each step is as follows: Step 1: Defect Area Removal The purpose of this step is to remove defective regions from the image to be inspected, eliminate the interference of these defective regions on fiber uniformity analysis, and avoid abnormal data affecting the evaluation results. The specific implementation process is as follows: like Figure 2A As shown, images of the paper to be inspected are obtained by taking pictures with a mobile phone, a high-speed camera, or an online defect monitoring camera. In the original image containing defects (such as impurities, damage, local thickness abnormalities, etc.), the user selects the defect area by using the box annotation tool in the interactive operation interface. The system automatically records the coordinate parameters of the box, and the defect area will be excluded in the subsequent uniformity quantification analysis.

[0048] like Figure 2B As shown, based on the marked boxes, the user confirms the defect selection area using the interactive cropping tool. The interface will simultaneously display operation prompts. After user confirmation, the system performs a defect area removal operation: separating the defect area within the marked boxes from the original image, and simultaneously marking the corresponding location in the original image with a box to clearly indicate that this area will not participate in subsequent calculations. The image after defect removal is shown below. Figure 2C As shown.

[0049] In addition, the system will automatically save the complete image file after removing defects. The file format is compatible with common image formats (such as PNG and JPG), and the resolution is consistent with the original image (default 300dpi). This provides defect-free basic data for subsequent partitioning and cutting steps, and saves the coordinate log of the defect area for easy traceability and verification later.

[0050] Step 2: Image partitioning and cropping

[0051] The purpose of this step is to divide the effective image after defect removal into independent sub-regions, ensuring the independence and comprehensiveness of the analysis of each region, so as to achieve refined analysis of uniformity. The specific implementation process is as follows: like Figure 2C As shown, the system reads the defect-free valid image saved in step 1. Based on the image's pixel coordinate system, and taking the smallest bounding rectangle of the valid area as the reference, it automatically divides the valid area into four non-overlapping sub-regions: Top, Bottom, Left, and Right, along the horizontal central axis (1 / 2 of the rectangle's width) and the vertical central axis (1 / 2 of the rectangle's height).

[0052] The boundaries of each sub-region are defined as follows: horizontally, the boundary is half the width of the effective region (pixel coordinates), and vertically, the boundary is half the height of the effective region (pixel coordinates). The partition boundaries are clear and there are no pixel repetitions or omissions. If the effective region is irregularly shaped (e.g., due to partial edge loss caused by defect removal), the system adjusts the cutting boundary using a pixel filling judgment algorithm to ensure that the effective pixel area ratio of each sub-region is ≥20%, thus avoiding analysis distortion caused by sub-regions being too small. If the effective pixel area ratio of a certain sub-region is <20%, the region is merged into the adjacent sub-region with the largest area, and the boundary and effective pixel area of ​​the merged sub-region are recalculated (to avoid analysis distortion caused by sub-regions being too small). After the segmentation is completed, the system automatically saves the image data of each sub-region and records the effective pixel area of ​​each sub-region (denoted as S1, S2, S3, and S4, respectively, corresponding to...). Figure 2D The upper area shown Figure 2E The lower area shown Figure 2F The left area shown Figure 2G (The right-hand area shown).

[0053] Step 3: Partition Weighted Analysis

[0054] The purpose of this step is to quantify the fiber uniformity of each sub-region and to set weights based on the sub-region area in order to improve the objectivity of the analysis results. The specific implementation process is as follows: Weight calculation: The area ratio (number of pixels) of each partition is calculated by the following mathematical formula (1), which is used as the weight of each partition.

[0055] Mathematical formula (1)

[0056] Among them, Weight i This represents the weight of the i-th partition (i=1 corresponds to the upper region, i=2 to the lower region, i=3 to the left region, and i=4 to the right region), Areai ∑Area represents the effective pixel area of ​​the i-th partition (i.e., S1, S2, S3, and S4 recorded in step 2). all parts This represents the total effective pixel area of ​​the four partitions.

[0057] Multi-dimensional uniformity index analysis: For each partition, a uniformity analysis is performed separately, and each index is output. The index parameters for each partition are as follows: Figure 3A As shown: 1) Average gray value (Mean): As the basic parameter for subsequent calculation of global uniformity, standard uniformity index and other indicators, its value reflects the overall brightness of fiber distribution in the partition (indirectly reflecting the fiber packing density in the area), and is calculated by the following mathematical formula (2): Mathematical expression (2) in, This represents the average grayscale value of the i-th partition (corresponding to the Top / Bottom / Left / Right regions divided in step 2). Let be the pixel height (row number) of the i-th partition. Let be the pixel width (number of columns) of the i-th partition. This represents the total pixel area of ​​the partition. This represents the grayscale value of the pixel at coordinates (x, y) within the i-th partition.

[0058] 2) Gray Standard Deviation (Std): The standard deviation of gray values ​​in the calculated area reflects the dispersion of gray values. The smaller the standard deviation, the more uniform the fiber distribution in the area. It is calculated using the following mathematical formula (3): Mathematical expression (3) in, This represents the global uniformity (grayscale standard deviation) of the i-th partition. Let be the pixel height (row number) of the i-th partition. Let be the pixel width (number of columns) of the i-th partition. This represents the total pixel area of ​​the partition. This represents the pixel grayscale value at coordinates (x, y) within the i-th partition. Let be the average gray value of the i-th partition.

[0059] 3) Mean Gradient: The gradient value reflects the intensity of pixel grayscale changes, corresponding to detailed features such as fiber edges and texture transitions.

[0060] This index is calculated by taking the arithmetic mean of the gradient values ​​of all pixels within a partition, reflecting the average intensity of fiber texture changes within that partition: the larger the average gradient value, the clearer the fiber edges and the more drastic the texture changes (there may be irregular fiber arrangement); the smaller the average gradient value, the smoother the texture changes and the more regular the fiber arrangement. It is calculated using the following mathematical formula (4): Mathematical expression (4) in, This represents the mean gradient of the i-th partition. Let be the pixel height (row number) of the i-th partition. Let be the pixel width (number of columns) of the i-th partition. G(x,y) represents the total pixel area of ​​the partition, and G(x,y) represents the gradient value of each pixel.

[0061] 4) Gray Standard Deviation Index (Std Exponential): The standard deviation index is a "relative uniformity" index derived from the gray standard deviation. Its core is to eliminate the interference caused by the difference in average gray values ​​(avoiding misjudgment of uniformity due to the different fiber base densities in different zones), making the index more sensitive to uniformity differences. It is calculated by the following mathematical formula (5): Mathematical expression (5) in, This represents the standard deviation index of the i-th partition. This represents the standard deviation of the grayscale value of the i-th partition. Let be the average gray value of the i-th partition.

[0062] 5) Gradient Exponential: The gradient exponent is a core indicator for quantifying "texture uniformity". It highlights the differences in texture consistency by exponentializing the dispersion of the gradient mean or gradient distribution.

[0063] The smaller the value, the more uniform the fiber edge variation within the partition, with no local texture abrupt changes, edge disorder, or other abnormalities; the larger the value, the more irregular the local texture (such as dense edges caused by fiber agglomeration). The calculation is performed using the following mathematical formulas (6) and (7): Mathematical expression (6) Mathematical expression (7) in, This represents the gradient exponent of the i-th partition. This represents the mean gradient of the i-th partition. Let represent the gradient standard deviation of the i-th partition.

[0064] Step 4: Overall weighted results

[0065] The purpose of this step is to integrate the weighted indices of each sub-region and output the final comprehensive result of fiber distribution uniformity. The specific implementation process is as follows: The weighted comprehensive calculation of each quantitative indicator is performed using the following mathematical formulas (7) to (11), where, This represents the weight of the i-th sub-region.

[0066] Overall average grayscale value : Mathematical expression (7) Overall grayscale standard deviation : Mathematical expression (8) Overall grayscale gradient mean : Mathematical expression (9) Overall gray standard deviation index : Mathematical expression (10) Overall grayscale gradient index : Mathematical expression (11) Ultimately, as Figure 3B As shown, the system outputs a test report containing "quantitative indicators, weight coefficients, and comprehensive indicators for each sub-region".

[0067] in, A smaller value indicates a lower overall gray level for the entire sheet of defect-free paper. The smaller the value, the more uniform the fiber density distribution. A lower value indicates better relative uniformity after eliminating the interference of differences in fiber base density. The smaller the value, the more consistent the fiber texture distribution, with no abnormalities such as local texture abrupt changes or messy edges (such as the problem of dense edges caused by fiber agglomeration).

[0068] Finally, refer to Figures 4A to 4D The reliability of the gradient index obtained by measuring the paper fiber using the paper fiber measurement method of the present invention is verified.

[0069] like Figure 4A As shown, the experiment selected a total of 23 test samples with sample numbers 1-23, covering paper types in typical application scenarios such as printing paper and packaging paper, which can effectively ensure the comprehensiveness and representativeness of the verification experiment.

[0070] The same production process parameters were collected synchronously for each test sample, providing basic data for subsequent analysis of the "correlation between process parameters and fiber uniformity".

[0071] like Figure 4B As shown, the table is a comparison table of core test data for 23 samples, including sample number, "online measurement gradient" value measured by the method of this invention, and "laboratory measurement parameter" value (i.e., industry-recognized fiber uniformity benchmark value) measured by standard laboratory methods (e.g., uniformity tester method).

[0072] like Figure 4C As shown, this figure is a comparison curve between "online gradient measurement value" and "laboratory measurement parameter value". The horizontal axis represents the sample number (1-23), and the vertical axis represents the index value (range 0-180).

[0073] like Figure 4D As shown in the figure, this is the fitting curve of the two, with the fitting equation being y = -1.0675x + 21143, where x is the online measured gradient value and y is the laboratory measured parameter value, and the coefficient of determination R² = 0.8027. The negative correlation fitting curve conforms to the technical logic—the higher the laboratory measured parameter value (the better the uniformity determined by the laboratory), the lower the gradient online measured value (the more regular the fiber texture determined by this invention). R² = 0.8027 is close to 1, indicating a strong linear correlation between the two sets of data, confirming that the gradient index measured by this invention can accurately reflect the uniformity of fiber distribution, and the detection results are reliable. It can effectively replace traditional laboratory testing and is suitable for the online quality control needs of papermaking production lines.

[0074] Furthermore, the present invention also provides a paper fiber measuring device, which includes a defect area removal unit, an image partitioning and cutting unit, a partition weighted analysis unit, and a comprehensive weighted result unit. The functions and roles of each part are implemented collaboratively with reference to the corresponding steps of the paper fiber measuring method described above, as follows: The aforementioned defect area removal unit is used to perform defect area removal operations: it receives user instructions to select and confirm defect areas in the original paper image through an interactive interface, records the coordinate parameters of the defect area, separates the area from the original image and marks it with a box, and simultaneously saves the effective image and coordinate log after defect removal. The image partitioning and cutting unit described above is used to perform image partitioning and cutting operations: read the effective image after removing defects, and automatically divide it into four non-overlapping sub-regions—upper, lower, left, and right—along the horizontal and vertical central axes based on the pixel coordinate system and the minimum bounding rectangle reference, and record the effective pixel area of ​​each sub-region. The aforementioned partition weighted analysis unit is used to perform partition weighted analysis operations: it calculates the area ratio as the weight based on the effective pixel area of ​​each sub-region, and then performs multi-dimensional uniformity index analysis on each sub-region, outputting quantitative parameters such as gray-level average value, gray-level standard deviation, gray-level gradient mean, gray-level standard deviation index, and gray-level gradient index. Each index is calculated using the corresponding mathematical formula. The aforementioned comprehensive weighted result section is used to perform comprehensive weighted result operations: based on the weight coefficients of each sub-region, the above multi-dimensional quantitative indicators are weighted and summed respectively, and a test report containing the quantitative indicators of each sub-region, weight coefficients, and comprehensive indicators (comprehensive global gray average value, gray standard deviation, gray gradient mean, gray standard deviation index, and gray gradient index) is output, which can effectively reflect the uniformity of paper fiber distribution.

[0075] <Variation Example>

[0076] Example 1: Lightweight Solution for Mobile Devices

[0077] This variant is designed for scenarios such as offline sampling inspection and rapid on-site verification. It is a lightweight adaptation of the original technical solution, with the key being to retain the uniformity quantification analysis logic and simplify hardware dependencies and operation processes.

[0078] Specifically, a web front-end interface based on Streamlit (a Python-based web application framework) was developed, supporting access from multiple terminals such as mobile phones, tablets, and laptops. No dedicated client installation is required; detection operations can be completed through a browser.

[0079] For example, by optimizing the image acquisition module and adapting it to mobile device cameras (supporting 1080P and above resolution shooting), and with built-in image preprocessing functions (automatic tilt correction and contrast enhancement), it can meet the image quality requirements of non-professional shooting scenarios. The zoning and weighting logic is simplified, and to address the computing power limitations of mobile devices, it offers a "Fast Mode" (dividing only 2 symmetrical sub-regions) and a "Precise Mode" (retaining 4 zones). The weight calculation uses a simplified algorithm to ensure second-level output of detection results. Core functions retain support for manually selecting defect areas, displaying multi-dimensional uniformity indicators (grayscale mean, standard deviation, gradient exponent), and exporting inspection reports (compatible with PDF / Excel formats), enabling rapid verification of paper fiber uniformity anytime, anywhere.

[0080] Example 2: Automatic Defect Detection and Cutting Solution

[0081] This variation introduces an automated algorithm to adapt to the online inspection needs of high-speed paper production lines, reducing manual intervention and improving inspection efficiency and consistency.

[0082] Integrating deep learning-based paper defect detection models (such as deep learning-based defect localization models), the system performs millisecond-level analysis on paper images acquired in real time from the production line. It automatically identifies common defects such as holes, stains, wrinkles, and fiber aggregates, accurately marking the coordinates of defect areas without the need for manual selection. The system dynamically adjusts its sub-region division strategy based on the location, size, and distribution of defects. If defects are concentrated on one side, it automatically optimizes the position of the central axis to ensure that the remaining effective area is evenly divided into four sub-regions, avoiding analysis distortion caused by defect distribution.

[0083] Based on the dynamically divided sub-regions, the system automatically calculates the effective pixel area of ​​each region, calculates the weight coefficients using mathematical formula (1), and then sequentially completes the quantification and weighted comprehensive calculation of multi-dimensional indicators, all without manual operation. This solution is suitable for high-speed paper production lines with speeds of several meters per second, enabling fully automatic online detection of fiber uniformity. The detection results are fed back to the production control system in real time, providing immediate data support for process adjustments, while reducing labor costs and scrap rates.

[0084] (Technical effect)

[0085] The paper fiber measurement method and device of this invention can accurately remove defective areas such as holes, stains, and wrinkles from paper images, completely eliminating the interference of defects on uniformity analysis and solving the core problem of distorted quantitative indicators caused by defects in traditional testing. By utilizing zoned area weighting and multi-dimensional quantitative analysis such as grayscale mean, standard deviation, and gradient index, a refined and objective assessment of fiber distribution uniformity is achieved. The coefficient of determination (R²) between the comprehensive indicators and laboratory standard test results reaches 0.8027, demonstrating high reliability and accuracy, and effectively replacing traditional offline testing methods.

[0086] Furthermore, this method and apparatus are adaptable to various scenarios, including online inspection and offline sampling, in high-speed paper production lines. They support automated defect identification and dynamic zoning adjustments (adapting to production speeds of several meters per second), and can also meet the rapid verification needs of mobile devices through a lightweight design, significantly improving inspection efficiency and reducing labor costs and scrap rates. The comprehensive quantitative indicators output provide precise data support for production process adjustments, helping to optimize key properties such as paper mechanical strength and printability, ensuring product quality stability. It is particularly suitable for quality control of various types of paper, including printing paper, packaging paper, and specialty industrial paper, meeting the requirements of high-speed, wide-format, and high-quality development in the modern paper industry.

[0087] While the structure and working principle of the present invention have been described above in conjunction with preferred embodiments, those skilled in the art should recognize that the above examples are merely illustrative and do not constitute a limitation of the invention. Modifications and variations can be made to the invention within the spirit and scope of the claims, and all such modifications and variations will fall within the protection scope of the present invention.

Claims

1. A method for measuring paper fibers, used to quantitatively detect the uniformity of paper fiber distribution, characterized in that, The following steps are performed sequentially: Step 1: Defect area removal. Obtain the paper image to be inspected, remove the defective areas in the paper image and save the defect-free valid image; Step 2: Image partitioning and segmentation, dividing the defect-free valid image into multiple non-overlapping sub-regions and recording the effective pixel area of ​​each sub-region; Step 3: Partition weighted analysis. Calculate the weight based on the effective pixel area of ​​each sub-region and perform multi-dimensional uniformity index analysis on each sub-region. Step 4: Based on the weighted results, the multi-dimensional uniformity index of each sub-region is weighted and calculated to output the comprehensive detection data of fiber distribution uniformity.

2. The paper fiber measurement method as described in claim 1, characterized in that, In step 1, the user's instruction to select the defective area in the original paper image is received through the interactive operation interface. The defective area is separated from the original image and marked with a box. The effective image after removing the defect is saved simultaneously.

3. The paper fiber measurement method as described in claim 1 or 2, characterized in that, In step 2, based on the pixel coordinate system, the area is divided into four sub-regions—upper, lower, left, and right—along the horizontal and vertical central axes, with the minimum bounding rectangle of the effective area as the reference.

4. The paper fiber measurement method as described in claim 1, characterized in that, In step 3, the weights are calculated using the following mathematical formula: Among them, Weight i Area represents the weight of the i-th partition. i Let ∑Area represent the effective pixel area of ​​the i-th partition. allparts This represents the total effective pixel area of ​​all sub-regions.

5. The paper fiber measurement method as described in claim 1, characterized in that, In step 3, the multi-dimensional uniformity index includes the average grayscale value, which is calculated using the following mathematical formula: in, This represents the average grayscale value of the i-th partition. Let be the pixel height (row number) of the i-th partition. Let be the pixel width (number of columns) of the i-th partition. This represents the total pixel area of ​​the partition. This represents the grayscale value of the pixel at coordinates (x, y) within the i-th partition.

6. The paper fiber measurement method as described in claim 5, characterized in that, In step 3, the multidimensional uniformity index includes the grayscale standard deviation, which is calculated using the following mathematical formula: in, This represents the global uniformity (grayscale standard deviation) of the i-th partition. Let be the pixel height (row number) of the i-th partition. Let be the pixel width (number of columns) of the i-th partition. This represents the total pixel area of ​​the partition. This represents the pixel grayscale value at coordinates (x, y) within the i-th partition. Let be the average gray value of the i-th partition.

7. The paper fiber measurement method as described in claim 6, characterized in that, In step 3, the multi-dimensional uniformity index includes the mean gray-level gradient, which is calculated using the following mathematical formula: in, This represents the mean gradient of the i-th partition. Let be the pixel height (row number) of the i-th partition. Let be the pixel width (number of columns) of the i-th partition. G(x,y) represents the total pixel area of ​​the partition, and G(x,y) represents the gradient value of each pixel.

8. The paper fiber measurement method as described in claim 7, characterized in that, In step 3, the multidimensional uniformity index further includes the gray standard deviation index calculated based on the gray standard deviation and the gray average value, and the gray gradient index calculated based on the gray gradient mean and the gray standard deviation.

9. The paper fiber measurement method as described in claim 8, characterized in that, In step 4, the comprehensive weighted calculation includes at least one of the following: the overall global grayscale average value, grayscale standard deviation, grayscale gradient mean, grayscale standard deviation index, and grayscale gradient index.

10. The paper fiber measuring device as claimed in claim 1, characterized in that, Methods for acquiring the paper image include: mobile phone shooting, high-speed camera shooting, and online defect monitoring camera shooting.

11. A paper fiber measuring device for implementing the paper fiber measuring method of claim 1, characterized in that, The paper fiber measuring device includes: A defect region removal unit is used to perform defect region removal operations and save defect-free valid images; An image partitioning and cutting unit is used to perform image partitioning and cutting operations and record the effective pixel area of ​​each sub-region. The partition weighted analysis unit is used to calculate the weight of each sub-region and perform multi-dimensional uniformity index analysis. The comprehensive weighted result unit is used to perform weighted calculations on multi-dimensional uniformity indicators and output a test report.

12. The paper fiber measuring device as described in claim 11, characterized in that, The paper fiber measuring device integrates a paper defect detection model based on deep learning, which can automatically identify defective areas and mark their coordinates.

13. The paper fiber measuring device as described in claim 11, characterized in that, The paper fiber measuring device is based on a web-based front-end interface, adapted to mobile terminals, and provides a partition cutting function with selectable fast and precise modes.

Images